Identifying discriminative features from high-dimensional data using support vector machines